The search bar is the field on your store where a visitor types what they are looking for and gets a list of matching products. On-site search is the whole system behind it: the index of your products, the matching rules, the results page and the way it handles typos, synonyms and empty results.

It concerns any seller whose catalog is big enough that browsing takes effort, roughly 20 products and up. A creator with four digital products does not need one. A brand with 60 items in five collections loses sales without one, because the visitor who wants "linen shirt blue" does not want to open three collections to find it.

A search bar is the visible input, usually in the header of the storefront, with a magnifying glass icon. On-site search is what happens after the visitor types: the query is matched against product names, descriptions, tags, variants and sometimes collections, and a results page shows the products that fit, ranked by relevance. Good search also suggests results while the visitor types, corrects spelling, understands synonyms ("hoodie" and "sweatshirt"), and offers filters on the results page.

It is not the search engine that brings visitors to your site from the web; that is Google or Bing, and it is a matter of SEO. On-site search only looks inside your own catalog. It is also not the same as filters on a collection page, although the two work well together: filters narrow a list, search starts from a word.

Related vocabulary: autocomplete or search suggestions (results shown while typing), zero-results query (a search that finds nothing), synonym (two words that should match the same products), search index (the prepared copy of your catalog that search reads), query (what the visitor typed).

Why it matters

Visitors who search know what they want. That is why they convert at two to three times the rate of visitors who browse. Give them a working search bar and you turn intent into orders in one step. Give them a broken one, and you lose your best visitors at the exact moment they were ready to buy.

A worked example. A store with 80 products gets 5,000 visits a month and converts at 2%, or 100 orders at $48, which is $4,800. Suppose 12% of visitors use search when it is offered, so 600 searchers. Searchers convert at 5%, so 30 of those orders come from search. Now the search misses obvious queries: "tshirt" finds nothing because products are called "T-shirt", and "gift" returns an empty page. A third of searches end with zero results and those visitors leave. Fixing synonyms and typos brings searcher conversion from 5% to 7%, so 42 orders from search instead of 30. That is 12 more orders and $576 a month, from a configuration change that takes an afternoon.

Search also tells you what people want. The list of queries with no results is the cheapest product research you will ever get.

How it works

On-site search runs on a simple pipeline. Here is what happens between the keystroke and the results:

  • Indexing. Your catalog (names, descriptions, tags, variant names, SKUs, collections) is copied into a search index, updated whenever a product changes. Only products that are published and in stock, or marked as visible, should be indexed.
  • Query handling. The visitor's text is cleaned up: lowercase, accents removed, plural and singular treated as one, and obvious typos corrected against the words in the index.
  • Matching. The query is compared with the indexed fields. A match in the product name weighs more than a match in the description. Synonyms you have defined are expanded ("sneakers" also matches "trainers").
  • Ranking. Matching products are ordered. Exact name matches first, then partial matches, with a boost for products that sell well or are in stock.
  • Suggestions. While typing, the top three to five matches and popular searches appear under the field, so many visitors never reach the results page.
  • Results page. Products in a grid with photo, name and price, plus filters (price, category, size, color) and a sort option.
  • Zero results. When nothing matches, the page should say so clearly and show alternatives: popular products, a category list or a "did you mean".

Everything in this pipeline can be tuned by the seller: which fields count, which synonyms exist, which products are boosted.

Benchmarks and examples

Numbers to expect once search is in place:

  • Share of visitors who search: 8% to 15% on small stores, more on stores with large catalogs or a visible bar on mobile.
  • Conversion rate of searchers: 4% to 8%, against 1% to 3% for browsers.
  • Share of revenue from search: 15% to 30% on stores past 50 products.
  • Zero-results rate: under 5% of searches is healthy. Above 15% means your synonyms and product naming need work.
  • Searches per session: 1.5 to 2.5. Much higher means visitors are not finding what they typed.

Typical situations:

  • A creator with 8 digital products. No search needed. A clear list on the link-in-bio page does the job better.
  • A brand with 45 physical products in 6 collections. A search bar in the header, autocomplete, and five synonyms cover 90% of queries. Search is used by 10% of visitors and brings 20% of orders.
  • A store with 300 products and many variants. Search with filters becomes the main navigation. Product naming, tags and synonyms need a monthly review.

Common mistakes

  • Hiding the bar behind an icon on mobile. If the visitor has to find the magnifying glass, tap it, then wait for a field to appear, half of them will not.
  • Searching names only. "Waterproof" should find the jacket whose name is "Ridge Shell" because the description says waterproof.
  • Ignoring zero-results queries. They are a to-do list: add synonyms, rename products, or stock what people ask for.
  • Returning out-of-stock and hidden products. A visitor who finds the perfect item and then reads "sold out" is frustrated twice.
  • Dead-end empty page. "No results" with nothing else is an exit. Show popular products and categories instead.

Best practices

  • Show the bar in the header, open and visible, on mobile and desktop. A placeholder like "Search products" helps.
  • Index everything useful: name, description, tags, variant names, collections and SKUs. A customer with a reorder in mind may type the SKU from their last invoice.
  • Add synonyms from your own queries. After the first month, read the search log and map the words customers use to the words you use.
  • Turn on typo tolerance and autocomplete. Most mobile queries contain a typo. Suggestions while typing cut the effort to two or three characters.
  • Design the zero-results page. A short message, the top products, the categories and a contact link. Never a blank page.
  • Put filters on the results page once you pass 50 products: price, category, size, color, availability.
  • Review the search report monthly. Top queries, zero-results queries, conversion of searchers. Fix the top three problems each time.

In Roctify

Roctify's storefront is built on one shared catalog where every product carries its name, description, variants, SKUs, stock and price, so what a visitor finds is always the live product with current availability. The storefront lets visitors browse collections and product pages on mobile and desktop, with multi-language and multi-currency settings so product names and prices display in the visitor's language and currency. As a store grows past a few dozen products, clean product names, complete descriptions and consistent variant naming in the catalog are what make products easy to find on the storefront, and they also make the product data ready for any channel that uses it.

Once a visitor lands on the product they wanted, the path is the same on every Roctify channel: one-page checkout, Stripe, PayPal or cash on delivery, and 0% transaction fees on every plan.

FAQ

Around 20 products, or earlier if your products have many variants or technical names. Under that, a well-organised link-in-bio page or a single collection page is faster for the visitor than typing. Above 50 products, search with filters becomes the main way people navigate.

What should the search bar match on?

At minimum product names and descriptions. Add tags, variant names (colors, sizes), collections and SKUs. The more fields you index, the more natural language works: "blue linen shirt" should find the product whether "blue" is in the name, a variant or a tag.

How do I know if my search is working?

Look at three numbers every month: the share of visitors who search, their conversion rate compared with non-searchers, and the share of searches with zero results. Searchers should convert at least twice as well as browsers, and zero-results searches should be under 5%. The zero-results list itself tells you exactly what to fix.